P985: the allelic ratio of driver and asxl1 mutations is prognostically relevant in pmf

Background: Primary myelofibrosis (PMF) is a clonal disorder driven by mutations in JAK2, CALR, and MPL. Somatic mutations in myeloid-associated genes were shown to impact prognosis of PMF patients (pts). Among these, ASXL1 mutations (ASXL1mt) are by far the most frequent and prognostically relevant, in fact they are included in the category of “high molecular risk” (HMR), along with EZH2, IDH1/2, SRSF2, and U2AF1Q157 mutations. Aims: To investigate the phenotypic and prognostic implications of ASXL1mt and variant allele frequency (VAF) in PMF. Methods: After IRB approval, consecutive pts with WHO-defined PMF were included in the study. Mutational analysis by targeted NGS was performed by previously described methods (Guglielmelli P et al., JCO 2017). Results: The study enrolled a total of 384 pts, including 190 (49%) prefibrotic and 194 (51%) overt PMF. Median age was 60 (18-90) years, 236 (61%) were male. JAK2, CALR and MPL mutations were found in 255 (66%), 83 (22%), and 17 (4%) pts, respectively; 36 (9%) were triple negative (TN). Among HMR mutations, ASXL1 was mutated in 88 (23%) pts, SRSF2 in 35 (9%), EZH2 in 32 (8%), U2AF1 in 15 (5%), and IDH1/2 in 11 (3%). VAF distribution of driver and HMR mutations was as reported in Fig.1A. We did not find any significant correlation of ASXL1mt VAF with other clinical and molecular characteristics, including PMF subtype, gender, age, hemoglobin, leukocyte, platelet and blast counts, splenomegaly, BM fibrosis grade, cytogenetics, mutational status for driver and other HMR genes and their VAFs. The only exception was the association of CALR-mutated PMF with lower ASXL1 VAF (median 28% vs 40%, p=.0332). Median overall survival (OS) was 120 (102-151) months. In univariate Cox analysis, ASXL1mt VAF did not correlate with OS, nor did other single driver and HMR mutations. When considered as a whole (for the purpose of this study, TN pts were considered has having driver VAF equal to 0), the VAF of driver mutations inversely correlated with survival (HR 0.4 [0.2-0.9], p=.0219). We next investigated the allelic ratio between driver and ASXL1 mutations (d/Aratio) and its prognostic correlates. Median (range) of d/Aratio was 1.24 (0-13), and ROC analysis with death as an endpoint identified 1.31 as the optimal cut-off value. In univariate analysis, pts with a d/Aratio <1.31 had a significantly worse OS compared to those with a ratio ≥1.31 (median 45 vs 104 months, p=.0014; HR 2.4 [1.4-4.1]) (Fig.1B). When TN pts with ASXL1mt were considered apart, having a d/Aratio <1.31 retained its inferior prognostic impact compared to the d/Aratio ≥1.31 group (median OS 57 vs 104 months, p<0.0453; HR 1.8 [1-3.2]), with TN/ASXL1mt pts being associated with the worst outcome (median OS 24 months) (Fig.1C). In a multivariate Cox model including VAF-adjusted ASXL1 mutant status (ASXL1wtvs ASXL1mt with d/Aratio ≥1.3 vs ASXL1mt with d/Aratio <1.3) and other HMR mutations, the former and mutated SRSF2 were confirmed to be independent predictors of inferior OS. Both d/Aratio <1.31 and ≥1.31 remained significant compared to ASXL1wt with respective HRs of 3.4 (2.3-5.1; p